Demand forecasting
Forecast Bias
By the Supply Chain Manager team · Updated October 16, 2026
Forecast bias is a consistent tendency to forecast too high or too low, measured by whether errors lean in one direction over time.
Accuracy measures how big errors are; bias measures which way they lean. A biased forecast does steady damage: always high builds overstock, always low causes repeated stockouts.
Bias often has a human source: sales teams inflating forecasts to protect availability, or a model that never learned demand fell after a big customer left.
Small bias is normal. What matters is whether it persists. A practical approach is to check bias by SKU group each month and look into any group that leans the same way for three months or more. Fixing bias often means correcting a data issue, such as a lost customer still in the history, rather than changing the model.
How to calculate forecast bias
Bias % = Σ (Forecast − Actual) ÷ Σ Actual × 100 Positive = over-forecasting; negative = under-forecasting
Forecast bias example
Illustrative numbers: over 12 weeks a SKU was forecast at 1,200 units and sold 1,000. Bias = +20%. Even if weekly errors look small, that is 200 units of stock you did not need.
Why forecast bias matters for distributors
Bias compounds into inventory problems faster than random error, because the errors do not cancel out. It is also usually the easiest problem to fix once it is visible.
Common forecast bias mistakes
- Watching only absolute error measures and missing a steady lean.
- Forecasting from shipments during stockouts, which creates a downward bias.
Examples on this page use made-up round numbers to show the calculation. They are not benchmarks or results from real businesses.